| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
As message in commit 7df676974359 ("mm/memory.c: Update local TLB if PTE
entry exists") said, we should update local TLB only on the second thread.
So in the do_anonymous_page() here, we should use update_mmu_tlb()
instead of update_mmu_cache() on the second thread.
As David pointed out, this is a performance improvement, not a
correctness fix.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Qi Zheng <[email protected]>
Reviewed-by: Muchun Song <[email protected]>
Acked-by: David Hildenbrand <[email protected]>
Cc: Bibo Mao <[email protected]>
Cc: Chris Zankel <[email protected]>
Cc: Huacai Chen <[email protected]>
Cc: Max Filippov <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
GCC's -Warray-bounds option detects out-of-bounds accesses to
statically-sized allocations in krealloc out-of-bounds tests.
Use OPTIMIZER_HIDE_VAR to suppress the warning.
Also change kmalloc_memmove_invalid_size to use OPTIMIZER_HIDE_VAR
instead of a volatile variable.
Link: https://lkml.kernel.org/r/e94399242d32e00bba6fd0d9ec4c897f188128e8.1664215688.git.andreyknvl@google.com
Signed-off-by: Andrey Konovalov <[email protected]>
Reported-by: kernel test robot <[email protected]>
Reviewed-by: Kees Cook <[email protected]>
Cc: Alexander Potapenko <[email protected]>
Cc: Andrey Ryabinin <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Marco Elver <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Link: https://lkml.kernel.org/r/a73cf109de0224cfd118d22be58ddebac3ae2897.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: Felix Kuehling <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Christian König <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Lyude Paul <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: Zi Yan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
When the module is unloaded or a GPU is unbound from the module it is
possible for device private pages to still be mapped in currently running
processes. This can lead to a hangs and RCU stall warnings when unbinding
the device as memunmap_pages() will wait in an uninterruptible state until
all device pages have been freed which may never happen.
Fix this by migrating device mappings back to normal CPU memory prior to
freeing the GPU memory chunks and associated device private pages.
Link: https://lkml.kernel.org/r/66277601fb8fda9af408b33da9887192bf895bda.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Cc: Lyude Paul <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: Christian König <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Felix Kuehling <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: Zi Yan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
nouveau_dmem_fault_copy_one() is used during handling of CPU faults via
the migrate_to_ram() callback and is used to copy data from GPU to CPU
memory. It is currently specific to fault handling, however a future
patch implementing eviction of data during teardown needs similar
functionality.
Refactor out the core functionality so that it is not specific to fault
handling.
Link: https://lkml.kernel.org/r/20573d7b4e641a78fde9935f948e64e71c9e709e.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Reviewed-by: Lyude Paul <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: Christian König <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Felix Kuehling <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: Zi Yan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Device drivers can use the migrate_vma family of functions to migrate
existing private anonymous mappings to device private pages. These pages
are backed by memory on the device with drivers being responsible for
copying data to and from device memory.
Device private pages are freed via the pgmap->page_free() callback when
they are unmapped and their refcount drops to zero. Alternatively they
may be freed indirectly via migration back to CPU memory in response to a
pgmap->migrate_to_ram() callback called whenever the CPU accesses an
address mapped to a device private page.
In other words drivers cannot control the lifetime of data allocated on
the devices and must wait until these pages are freed from userspace.
This causes issues when memory needs to reclaimed on the device, either
because the device is going away due to a ->release() callback or because
another user needs to use the memory.
Drivers could use the existing migrate_vma functions to migrate data off
the device. However this would require them to track the mappings of each
page which is both complicated and not always possible. Instead drivers
need to be able to migrate device pages directly so they can free up
device memory.
To allow that this patch introduces the migrate_device family of functions
which are functionally similar to migrate_vma but which skips the initial
lookup based on mapping.
Link: https://lkml.kernel.org/r/868116aab70b0c8ee467d62498bb2cf0ef907295.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Zi Yan <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Christian König <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: Felix Kuehling <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Lyude Paul <[email protected]>
Cc: Michael Ellerman <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
migrate_device_coherent_page() reuses the existing migrate_vma family of
functions to migrate a specific page without providing a valid mapping or
vma. This looks a bit odd because it means we are calling migrate_vma_*()
without setting a valid vma, however it was considered acceptable at the
time because the details were internal to migrate_device.c and there was
only a single user.
One of the reasons the details could be kept internal was that this was
strictly for migrating device coherent memory. Such memory can be copied
directly by the CPU without intervention from a driver. However this
isn't true for device private memory, and a future change requires similar
functionality for device private memory. So refactor the code into
something more sensible for migrating device memory without a vma.
Link: https://lkml.kernel.org/r/c7b2ff84e9b33d022cf4a40f87d051f281a16d8f.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Zi Yan <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Christian König <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: Felix Kuehling <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Lyude Paul <[email protected]>
Cc: Michael Ellerman <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
ZONE_DEVICE pages have a struct dev_pagemap which is allocated by a
driver. When the struct page is first allocated by the kernel in
memremap_pages() a reference is taken on the associated pagemap to ensure
it is not freed prior to the pages being freed.
Prior to 27674ef6c73f ("mm: remove the extra ZONE_DEVICE struct page
refcount") pages were considered free and returned to the driver when the
reference count dropped to one. However the pagemap reference was not
dropped until the page reference count hit zero. This would occur as part
of the final put_page() in memunmap_pages() which would wait for all pages
to be freed prior to returning.
When the extra refcount was removed the pagemap reference was no longer
being dropped in put_page(). Instead memunmap_pages() was changed to
explicitly drop the pagemap references. This means that memunmap_pages()
can complete even though pages are still mapped by the kernel which can
lead to kernel crashes, particularly if a driver frees the pagemap.
To fix this drivers should take a pagemap reference when allocating the
page. This reference can then be returned when the page is freed.
Link: https://lkml.kernel.org/r/12d155ec727935ebfbb4d639a03ab374917ea51b.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Fixes: 27674ef6c73f ("mm: remove the extra ZONE_DEVICE struct page refcount")
Cc: Jason Gunthorpe <[email protected]>
Cc: Felix Kuehling <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Christian König <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Lyude Paul <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: Zi Yan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Since 27674ef6c73f ("mm: remove the extra ZONE_DEVICE struct page
refcount") device private pages have no longer had an extra reference
count when the page is in use. However before handing them back to the
owning device driver we add an extra reference count such that free pages
have a reference count of one.
This makes it difficult to tell if a page is free or not because both free
and in use pages will have a non-zero refcount. Instead we should return
pages to the drivers page allocator with a zero reference count. Kernel
code can then safely use kernel functions such as get_page_unless_zero().
Link: https://lkml.kernel.org/r/cf70cf6f8c0bdb8aaebdbfb0d790aea4c683c3c6.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Acked-by: Felix Kuehling <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Christian König <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Lyude Paul <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: Zi Yan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Patch series "Fix several device private page reference counting issues",
v2
This series aims to fix a number of page reference counting issues in
drivers dealing with device private ZONE_DEVICE pages. These result in
use-after-free type bugs, either from accessing a struct page which no
longer exists because it has been removed or accessing fields within the
struct page which are no longer valid because the page has been freed.
During normal usage it is unlikely these will cause any problems. However
without these fixes it is possible to crash the kernel from userspace.
These crashes can be triggered either by unloading the kernel module or
unbinding the device from the driver prior to a userspace task exiting.
In modules such as Nouveau it is also possible to trigger some of these
issues by explicitly closing the device file-descriptor prior to the task
exiting and then accessing device private memory.
This involves some minor changes to both PowerPC and AMD GPU code.
Unfortunately I lack hardware to test either of those so any help there
would be appreciated. The changes mimic what is done in for both Nouveau
and hmm-tests though so I doubt they will cause problems.
This patch (of 8):
When the CPU tries to access a device private page the migrate_to_ram()
callback associated with the pgmap for the page is called. However no
reference is taken on the faulting page. Therefore a concurrent migration
of the device private page can free the page and possibly the underlying
pgmap. This results in a race which can crash the kernel due to the
migrate_to_ram() function pointer becoming invalid. It also means drivers
can't reliably read the zone_device_data field because the page may have
been freed with memunmap_pages().
Close the race by getting a reference on the page while holding the ptl to
ensure it has not been freed. Unfortunately the elevated reference count
will cause the migration required to handle the fault to fail. To avoid
this failure pass the faulting page into the migrate_vma functions so that
if an elevated reference count is found it can be checked to see if it's
expected or not.
[[email protected]: fix build]
Link: https://lkml.kernel.org/r/[email protected]
Link: https://lkml.kernel.org/r/cover.60659b549d8509ddecafad4f498ee7f03bb23c69.1664366292.git-series.apopple@nvidia.com
Link: https://lkml.kernel.org/r/d3e813178a59e565e8d78d9b9a4e2562f6494f90.1664366292.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <[email protected]>
Acked-by: Felix Kuehling <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Ralph Campbell <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Lyude Paul <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Alex Sierra <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Christian König <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: "Huang, Ying" <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Yang Shi <[email protected]>
Cc: Zi Yan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
In many places we can use damon_sz_region() to instead of "r->ar.end -
r->ar.start".
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Xin Hao <[email protected]>
Suggested-by: SeongJae Park <[email protected]>
Reviewed-by: SeongJae Park <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Rename sz_damon_region() to damon_sz_region(), and move it to
"include/linux/damon.h", because in many places, we can to use this func.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Xin Hao <[email protected]>
Suggested-by: SeongJae Park <[email protected]>
Reviewed-by: SeongJae Park <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
alloc_pages(), kmalloc() and vmalloc() are all memory allocation functions
which can return NULL when some internal memory failures happen. So it is
better to check the return of them to catch the failure in time for better
test them.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Xiaoke Wang <[email protected]>
Reviewed-by: Alexander Potapenko <[email protected]>
Cc: Andrey Konovalov <[email protected]>
Cc: Andrey Ryabinin <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Marco Elver <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
This is an optimization to reduce stackdepot pressure.
struct mmu_gather contains 7 1-bit fields packed into a 32-bit unsigned
int value. The remaining 25 bits remain uninitialized and are never used,
but KMSAN updates the origin for them in zap_pXX_range() in mm/memory.c,
thus creating very long origin chains. This is technically correct, but
consumes too much memory.
Unpoisoning the whole structure will prevent creating such chains.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Alexander Potapenko <[email protected]>
Acked-by: Marco Elver <[email protected]>
Cc: Alexander Viro <[email protected]>
Cc: Alexei Starovoitov <[email protected]>
Cc: Andrey Konovalov <[email protected]>
Cc: Andrey Konovalov <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: Eric Biggers <[email protected]>
Cc: Eric Dumazet <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Herbert Xu <[email protected]>
Cc: Ilya Leoshkevich <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jens Axboe <[email protected]>
Cc: Joonsoo Kim <[email protected]>
Cc: Kees Cook <[email protected]>
Cc: Liu Shixin <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Michael S. Tsirkin <[email protected]>
Cc: Pekka Enberg <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Petr Mladek <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Vasily Gorbik <[email protected]>
Cc: Vegard Nossum <[email protected]>
Cc: Vlastimil Babka <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
The recent change of page_cache_ra_unbounded() arguments was buggy in the
two callers, causing us to readahead the wrong pages. Move the definition
of ractl down to after the index is set correctly. This affected
performance on configurations that use fs-verity.
Link: https://lkml.kernel.org/r/[email protected]
Fixes: 73bb49da50cd ("mm/readahead: make page_cache_ra_unbounded take a readahead_control")
Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>
Reported-by: Jintao Yin <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Commit c462ac288f2c ("mm: Introduce arch_validate_flags()") added a late
check in mmap_region() to let architectures validate vm_flags. The check
needs to happen after calling ->mmap() as the flags can potentially be
modified during this callback.
If arch_validate_flags() check fails we unmap and free the vma. However,
the error path fails to undo the ->mmap() call that previously succeeded
and depending on the specific ->mmap() implementation this translates to
reference increments, memory allocations and other operations what will
not be cleaned up.
There are several places (mainly device drivers) where this is an issue.
However, one specific example is bpf_map_mmap() which keeps count of the
mappings in map->writecnt. The count is incremented on ->mmap() and then
decremented on vm_ops->close(). When arch_validate_flags() fails this
count is off since bpf_map_mmap_close() is never called.
One can reproduce this issue in arm64 devices with MTE support. Here the
vm_flags are checked to only allow VM_MTE if VM_MTE_ALLOWED has been set
previously. From userspace then is enough to pass the PROT_MTE flag to
mmap() syscall to trigger the arch_validate_flags() failure.
The following program reproduces this issue:
#include <stdio.h>
#include <unistd.h>
#include <linux/unistd.h>
#include <linux/bpf.h>
#include <sys/mman.h>
int main(void)
{
union bpf_attr attr = {
.map_type = BPF_MAP_TYPE_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(long long),
.max_entries = 256,
.map_flags = BPF_F_MMAPABLE,
};
int fd;
fd = syscall(__NR_bpf, BPF_MAP_CREATE, &attr, sizeof(attr));
mmap(NULL, 4096, PROT_WRITE | PROT_MTE, MAP_SHARED, fd, 0);
return 0;
}
By manually adding some log statements to the vm_ops callbacks we can
confirm that when passing PROT_MTE to mmap() the map->writecnt is off upon
->release():
With PROT_MTE flag:
root@debian:~# ./bpf-test
[ 111.263874] bpf_map_write_active_inc: map=9 writecnt=1
[ 111.288763] bpf_map_release: map=9 writecnt=1
Without PROT_MTE flag:
root@debian:~# ./bpf-test
[ 157.816912] bpf_map_write_active_inc: map=10 writecnt=1
[ 157.830442] bpf_map_write_active_dec: map=10 writecnt=0
[ 157.832396] bpf_map_release: map=10 writecnt=0
This patch fixes the above issue by calling vm_ops->close() when the
arch_validate_flags() check fails, after this we can proceed to unmap and
free the vma on the error path.
Link: https://lkml.kernel.org/r/[email protected]
Fixes: c462ac288f2c ("mm: Introduce arch_validate_flags()")
Signed-off-by: Carlos Llamas <[email protected]>
Reviewed-by: Catalin Marinas <[email protected]>
Acked-by: Andrii Nakryiko <[email protected]>
Reviewed-by: Liam Howlett <[email protected]>
Cc: Christian Brauner (Microsoft) <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Suren Baghdasaryan <[email protected]>
Cc: <[email protected]> [5.10+]
Signed-off-by: Andrew Morton <[email protected]>
|
|
FIELD_GET() must only be used with a mask that is a compile-time
constant. Mark the functions as __always_inline to avoid the problem.
Fixes: 21b688dabecb6a ("net: phy: micrel: Cable Diag feature for lan8814 phy")
Reported-by: kernel test robot <[email protected]>
Signed-off-by: Divya Koppera <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
A WARN_ON call trace would be triggered when 'ct(commit, alg=helper)'
applies on a confirmed connection:
WARNING: CPU: 0 PID: 1251 at net/netfilter/nf_conntrack_extend.c:98
RIP: 0010:nf_ct_ext_add+0x12d/0x150 [nf_conntrack]
Call Trace:
<TASK>
nf_ct_helper_ext_add+0x12/0x60 [nf_conntrack]
__nf_ct_try_assign_helper+0xc4/0x160 [nf_conntrack]
__ovs_ct_lookup+0x72e/0x780 [openvswitch]
ovs_ct_execute+0x1d8/0x920 [openvswitch]
do_execute_actions+0x4e6/0xb60 [openvswitch]
ovs_execute_actions+0x60/0x140 [openvswitch]
ovs_packet_cmd_execute+0x2ad/0x310 [openvswitch]
genl_family_rcv_msg_doit.isra.15+0x113/0x150
genl_rcv_msg+0xef/0x1f0
which can be reproduced with these OVS flows:
table=0, in_port=veth1,tcp,tcp_dst=2121,ct_state=-trk
actions=ct(commit, table=1)
table=1, in_port=veth1,tcp,tcp_dst=2121,ct_state=+trk+new
actions=ct(commit, alg=ftp),normal
The issue was introduced by commit 248d45f1e193 ("openvswitch: Allow
attaching helper in later commit") where it somehow removed the check
of nf_ct_is_confirmed before asigning the helper. This patch is to fix
it by bringing it back.
Fixes: 248d45f1e193 ("openvswitch: Allow attaching helper in later commit")
Reported-by: Pablo Neira Ayuso <[email protected]>
Signed-off-by: Xin Long <[email protected]>
Acked-by: Aaron Conole <[email protected]>
Tested-by: Aaron Conole <[email protected]>
Link: https://lore.kernel.org/r/c5c9092a22a2194650222bffaf786902613deb16.1665085502.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Kuniyuki Iwashima says:
====================
tcp/udp: Fix memory leaks and data races around IPV6_ADDRFORM.
This series fixes some memory leaks and data races caused in the
same scenario where one thread converts an IPv6 socket into IPv4
with IPV6_ADDRFORM and another accesses the socket concurrently.
v4: https://lore.kernel.org/netdev/[email protected]/
v3 (Resend): https://lore.kernel.org/netdev/[email protected]/
v3: https://lore.kernel.org/netdev/[email protected]/
v2: https://lore.kernel.org/netdev/[email protected]/
v1: https://lore.kernel.org/netdev/[email protected]/
====================
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
setsockopt(IPV6_ADDRFORM) and tcp_v6_connect() change icsk->icsk_af_ops
under lock_sock(), but tcp_(get|set)sockopt() read it locklessly. To
avoid load/store tearing, we need to add READ_ONCE() and WRITE_ONCE()
for the reads and writes.
Thanks to Eric Dumazet for providing the syzbot report:
BUG: KCSAN: data-race in tcp_setsockopt / tcp_v6_connect
write to 0xffff88813c624518 of 8 bytes by task 23936 on cpu 0:
tcp_v6_connect+0x5b3/0xce0 net/ipv6/tcp_ipv6.c:240
__inet_stream_connect+0x159/0x6d0 net/ipv4/af_inet.c:660
inet_stream_connect+0x44/0x70 net/ipv4/af_inet.c:724
__sys_connect_file net/socket.c:1976 [inline]
__sys_connect+0x197/0x1b0 net/socket.c:1993
__do_sys_connect net/socket.c:2003 [inline]
__se_sys_connect net/socket.c:2000 [inline]
__x64_sys_connect+0x3d/0x50 net/socket.c:2000
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x2b/0x70 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
read to 0xffff88813c624518 of 8 bytes by task 23937 on cpu 1:
tcp_setsockopt+0x147/0x1c80 net/ipv4/tcp.c:3789
sock_common_setsockopt+0x5d/0x70 net/core/sock.c:3585
__sys_setsockopt+0x212/0x2b0 net/socket.c:2252
__do_sys_setsockopt net/socket.c:2263 [inline]
__se_sys_setsockopt net/socket.c:2260 [inline]
__x64_sys_setsockopt+0x62/0x70 net/socket.c:2260
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x2b/0x70 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
value changed: 0xffffffff8539af68 -> 0xffffffff8539aff8
Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 23937 Comm: syz-executor.5 Not tainted
6.0.0-rc4-syzkaller-00331-g4ed9c1e971b1-dirty #0
Hardware name: Google Google Compute Engine/Google Compute Engine,
BIOS Google 08/26/2022
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot <[email protected]>
Reported-by: Eric Dumazet <[email protected]>
Signed-off-by: Kuniyuki Iwashima <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Commit 086d49058cd8 ("ipv6: annotate some data-races around sk->sk_prot")
fixed some data-races around sk->sk_prot but it was not enough.
Some functions in inet6_(stream|dgram)_ops still access sk->sk_prot
without lock_sock() or rtnl_lock(), so they need READ_ONCE() to avoid
load tearing.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Originally, inet6_sk(sk)->XXX were changed under lock_sock(), so we were
able to clean them up by calling inet6_destroy_sock() during the IPv6 ->
IPv4 conversion by IPV6_ADDRFORM. However, commit 03485f2adcde ("udpv6:
Add lockless sendmsg() support") added a lockless memory allocation path,
which could cause a memory leak:
setsockopt(IPV6_ADDRFORM) sendmsg()
+-----------------------+ +-------+
- do_ipv6_setsockopt(sk, ...) - udpv6_sendmsg(sk, ...)
- sockopt_lock_sock(sk) ^._ called via udpv6_prot
- lock_sock(sk) before WRITE_ONCE()
- WRITE_ONCE(sk->sk_prot, &tcp_prot)
- inet6_destroy_sock() - if (!corkreq)
- sockopt_release_sock(sk) - ip6_make_skb(sk, ...)
- release_sock(sk) ^._ lockless fast path for
the non-corking case
- __ip6_append_data(sk, ...)
- ipv6_local_rxpmtu(sk, ...)
- xchg(&np->rxpmtu, skb)
^._ rxpmtu is never freed.
- goto out_no_dst;
- lock_sock(sk)
For now, rxpmtu is only the case, but not to miss the future change
and a similar bug fixed in commit e27326009a3d ("net: ping6: Fix
memleak in ipv6_renew_options()."), let's set a new function to IPv6
sk->sk_destruct() and call inet6_cleanup_sock() there. Since the
conversion does not change sk->sk_destruct(), we can guarantee that
we can clean up IPv6 resources finally.
We can now remove all inet6_destroy_sock() calls from IPv6 protocol
specific ->destroy() functions, but such changes are invasive to
backport. So they can be posted as a follow-up later for net-next.
Fixes: 03485f2adcde ("udpv6: Add lockless sendmsg() support")
Signed-off-by: Kuniyuki Iwashima <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Commit 4b340ae20d0e ("IPv6: Complete IPV6_DONTFRAG support") forgot
to add a change to free inet6_sk(sk)->rxpmtu while converting an IPv6
socket into IPv4 with IPV6_ADDRFORM. After conversion, sk_prot is
changed to udp_prot and ->destroy() never cleans it up, resulting in
a memory leak.
This is due to the discrepancy between inet6_destroy_sock() and
IPV6_ADDRFORM, so let's call inet6_destroy_sock() from IPV6_ADDRFORM
to remove the difference.
However, this is not enough for now because rxpmtu can be changed
without lock_sock() after commit 03485f2adcde ("udpv6: Add lockless
sendmsg() support"). We will fix this case in the following patch.
Note we will rename inet6_destroy_sock() to inet6_cleanup_sock() and
remove unnecessary inet6_destroy_sock() calls in sk_prot->destroy()
in the future.
Fixes: 4b340ae20d0e ("IPv6: Complete IPV6_DONTFRAG support")
Signed-off-by: Kuniyuki Iwashima <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
syzbot reported a memory leak [0] related to IPV6_ADDRFORM.
The scenario is that while one thread is converting an IPv6 socket into
IPv4 with IPV6_ADDRFORM, another thread calls do_ipv6_setsockopt() and
allocates memory to inet6_sk(sk)->XXX after conversion.
Then, the converted sk with (tcp|udp)_prot never frees the IPv6 resources,
which inet6_destroy_sock() should have cleaned up.
setsockopt(IPV6_ADDRFORM) setsockopt(IPV6_DSTOPTS)
+-----------------------+ +----------------------+
- do_ipv6_setsockopt(sk, ...)
- sockopt_lock_sock(sk) - do_ipv6_setsockopt(sk, ...)
- lock_sock(sk) ^._ called via tcpv6_prot
- WRITE_ONCE(sk->sk_prot, &tcp_prot) before WRITE_ONCE()
- xchg(&np->opt, NULL)
- txopt_put(opt)
- sockopt_release_sock(sk)
- release_sock(sk) - sockopt_lock_sock(sk)
- lock_sock(sk)
- ipv6_set_opt_hdr(sk, ...)
- ipv6_update_options(sk, opt)
- xchg(&inet6_sk(sk)->opt, opt)
^._ opt is never freed.
- sockopt_release_sock(sk)
- release_sock(sk)
Since IPV6_DSTOPTS allocates options under lock_sock(), we can avoid this
memory leak by testing whether sk_family is changed by IPV6_ADDRFORM after
acquiring the lock.
This issue exists from the initial commit between IPV6_ADDRFORM and
IPV6_PKTOPTIONS.
[0]:
BUG: memory leak
unreferenced object 0xffff888009ab9f80 (size 96):
comm "syz-executor583", pid 328, jiffies 4294916198 (age 13.034s)
hex dump (first 32 bytes):
01 00 00 00 48 00 00 00 08 00 00 00 00 00 00 00 ....H...........
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<000000002ee98ae1>] kmalloc include/linux/slab.h:605 [inline]
[<000000002ee98ae1>] sock_kmalloc+0xb3/0x100 net/core/sock.c:2566
[<0000000065d7b698>] ipv6_renew_options+0x21e/0x10b0 net/ipv6/exthdrs.c:1318
[<00000000a8c756d7>] ipv6_set_opt_hdr net/ipv6/ipv6_sockglue.c:354 [inline]
[<00000000a8c756d7>] do_ipv6_setsockopt.constprop.0+0x28b7/0x4350 net/ipv6/ipv6_sockglue.c:668
[<000000002854d204>] ipv6_setsockopt+0xdf/0x190 net/ipv6/ipv6_sockglue.c:1021
[<00000000e69fdcf8>] tcp_setsockopt+0x13b/0x2620 net/ipv4/tcp.c:3789
[<0000000090da4b9b>] __sys_setsockopt+0x239/0x620 net/socket.c:2252
[<00000000b10d192f>] __do_sys_setsockopt net/socket.c:2263 [inline]
[<00000000b10d192f>] __se_sys_setsockopt net/socket.c:2260 [inline]
[<00000000b10d192f>] __x64_sys_setsockopt+0xbe/0x160 net/socket.c:2260
[<000000000a80d7aa>] do_syscall_x64 arch/x86/entry/common.c:50 [inline]
[<000000000a80d7aa>] do_syscall_64+0x38/0x90 arch/x86/entry/common.c:80
[<000000004562b5c6>] entry_SYSCALL_64_after_hwframe+0x63/0xcd
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot <[email protected]>
Signed-off-by: Kuniyuki Iwashima <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Conor Dooley <[email protected]> says:
From: Conor Dooley <[email protected]>
The device trees produced automatically for the virt and spike machines
fail dt-validate on several grounds. Some of these need to be fixed in
the linux kernel's dt-bindings, but others are caused by bugs in QEMU.
Patches been sent that fix the QEMU issues [0], but a couple of them
need to be fixed in the kernel's dt-bindings. The first patches add
compatibles for "riscv,{clint,plic}0" which are present in drivers and
the auto generated QEMU dtbs.
Thanks to Rob Herring for reporting these issues [1],
Conor.
To reproduce the errors:
./build/qemu-system-riscv64 -nographic -machine virt,dumpdtb=qemu.dtb
dt-validate -p /path/to/linux/kernel/Documentation/devicetree/bindings/processed-schema.json qemu.dtb
(The processed schema needs to be generated first)
0 - https://lore.kernel.org/linux-riscv/[email protected]/
1 - https://lore.kernel.org/linux-riscv/[email protected]/
* fix-dt-validate:
dt-bindings: riscv: add new riscv,isa strings for emulators
dt-bindings: interrupt-controller: sifive,plic: add legacy riscv compatible
dt-bindings: timer: sifive,clint: add legacy riscv compatible
Link: https://lore.kernel.org/r/[email protected]
[Palmer: some cover letter pruning, and dropped #4 as suggested.]
Signed-off-by: Palmer Dabbelt <[email protected]>
|
|
The QEMU virt and spike machines currently export a riscv,isa string of
"rv64imafdcsuh",
While the RISC-V foundation has been ratifying a bunch of extenstions
etc, the kernel has remained relatively static with what hardware is
supported - but the same is not true of QEMU. Using the virt machine
and running dt-validate on the dumped dtb fails, partly due to the
unexpected isa string.
Rather than enumerate the many many possbilities, change the pattern
to a regex, with the following assumptions:
- ima are required
- the single letter order is fixed & we don't care about things that
can't even do "ima"
- the standard multi letter extensions are all in a "_z<foo>" format
where the first letter of <foo> is a valid single letter extension
- _s & _h are used for supervisor and hyper visor extensions
- convention says that after the first two chars, a standard multi
letter extension name could be an english word (ifencei anyone?) so
it's not worth restricting the charset
- as the above is just convention, don't apply any charset restrictions
to reduce future churn
- vendor ISA extensions begind with _x and have no charset restrictions
- we don't care about an e extension from an OS pov
- that attempting to validate the contents of the multiletter extensions
with dt-validate beyond the formatting is a futile, massively verbose
or unwieldy exercise at best
The following limitations also apply:
- multi letter extension ordering is not enforced. dt-schema does not
appear to allow for named match groups, so the resulting regex would
be even more of a headache
- ditto for the numbered extensions
Finally, add me as a maintainer of the binding so that when it breaks
in the future, I can be held responsible!
Reported-by: Rob Herring <[email protected]>
Link: https://lore.kernel.org/linux-riscv/[email protected]/
Reviewed-by: Andrew Jones <[email protected]>
Acked-by: Guo Ren <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Acked-by: Heiko Stuebner <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
|
|
While "real" hardware might not use the compatible string "riscv,plic0"
it is present in the driver & QEMU uses it for automatically generated
virt machine dtbs. To avoid dt-validate problems with QEMU produced
dtbs, such as the following, add it to the binding.
riscv-virt.dtb: plic@c000000: compatible: 'oneOf' conditional failed, one must be fixed:
'sifive,plic-1.0.0' is not one of ['sifive,fu540-c000-plic', 'starfive,jh7100-plic', 'canaan,k210-plic']
'sifive,plic-1.0.0' is not one of ['allwinner,sun20i-d1-plic']
'sifive,plic-1.0.0' was expected
'thead,c900-plic' was expected
riscv-virt.dtb: plic@c000000: '#address-cells' is a required property
Reported-by: Rob Herring <[email protected]>
Link: https://lore.kernel.org/linux-riscv/[email protected]/
Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Reviewed-by: Heiko Stuebner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
|
|
While "real" hardware might not use the compatible string "riscv,clint0"
it is present in the driver & QEMU uses it for automatically generated
virt machine dtbs. To avoid dt-validate problems with QEMU produced
dtbs, such as the following, add it to the binding.
riscv-virt.dtb: clint@2000000: compatible:0: 'sifive,clint0' is not one of ['sifive,fu540-c000-clint', 'starfive,jh7100-clint', 'canaan,k210-clint']
Reported-by: Rob Herring <[email protected]>
Link: https://lore.kernel.org/linux-riscv/[email protected]/
Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Reviewed-by: Heiko Stuebner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>
|
|
When PTE_MARKER_UFFD_WP not configured, it's still possible to reach pte
marker code and trigger an warning. Add a few CONFIG_PTE_MARKER_UFFD_WP
ifdefs to make sure the code won't be reached when not compiled in.
Link: https://lkml.kernel.org/r/YzeR+R6b4bwBlBHh@x1n
Fixes: b1f9e876862d ("mm/uffd: enable write protection for shmem & hugetlbfs")
Signed-off-by: Peter Xu <[email protected]>
Reported-by: <[email protected]>
Cc: Axel Rasmussen <[email protected]>
Cc: Brian Geffon <[email protected]>
Cc: Edward Liaw <[email protected]>
Cc: Liu Shixin <[email protected]>
Cc: Mike Kravetz <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
If the brk VMA is the last vma in a maple node and meets the rare criteria
that it can be expanded, then preallocation is necessary to avoid a
potential fs_reclaim circular lock issue on low resources.
At the same time use the actual vma start address (unaligned) when calling
vma_adjust_trans_huge().
Link: https://lkml.kernel.org/r/[email protected]
Fixes: 2e7ce7d354f2 (mm/mmap: change do_brk_flags() to expand existing VMA and add do_brk_munmap())
Signed-off-by: Liam R. Howlett <[email protected]>
Reported-by: Yu Zhao <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Link: https://lkml.kernel.org/r/[email protected]
Fixes: 7964cf8caa4d ("mm: remove vmacache")
Signed-off-by: Alexey Dobriyan <[email protected]>
Acked-by: Liam Howlett <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
The anon vma was not unlinked and the file was not closed in the failure
path when the machine runs out of memory during the maple tree
modification. This caused a memory leak of the anon vma chain and vma
since neither would be freed.
Link: https://lkml.kernel.org/r/[email protected]
Fixes: 524e00b36e8c ("mm: remove rb tree")
Signed-off-by: Liam R. Howlett <[email protected]>
Reported-by: Lukas Bulwahn <[email protected]>
Tested-by: Lukas Bulwahn <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
When we successfully find a pageblock in fast_find_migrateblock(), the
block will be set skip-flag through set_pageblock_skip(). However, when
entering isolate_migratepages_block(), the whole pageblock will be skipped
due to the branch 'if (!valid_page && IS_ALIGNED(low_pfn,
pageblock_nr_pages))'. Eventually we will goto isolate_abort and isolate
nothing. That makes fast_find_migrateblock useless.
In this patch, when we find a suitable pageblock in
fast_find_migrateblock, we do noting but let isolate_migratepages_block to
set skip flag to the pageblock after scan it. Normally, we would isolate
some pages from the fast-find block.
I use mmtest/thpscale-madvhugepage test it. Here is the result:
baseline patch
Amean fault-both-1 1331.66 ( 0.00%) 1261.04 * 5.30%*
Amean fault-both-3 1383.95 ( 0.00%) 1191.69 * 13.89%*
Amean fault-both-5 1568.13 ( 0.00%) 1445.20 * 7.84%*
Amean fault-both-7 1819.62 ( 0.00%) 1555.13 * 14.54%*
Amean fault-both-12 1106.96 ( 0.00%) 1149.43 * -3.84%*
Amean fault-both-18 2196.93 ( 0.00%) 1875.77 * 14.62%*
Amean fault-both-24 2642.69 ( 0.00%) 2671.21 * -1.08%*
Amean fault-both-30 2901.89 ( 0.00%) 2857.32 * 1.54%*
Amean fault-both-32 3747.00 ( 0.00%) 3479.23 * 7.15%*
Link: https://lkml.kernel.org/r/[email protected]
Fixes: 70b44595eafe9 ("mm, compaction: use free lists to quickly locate a migration source")
Signed-off-by: zhouchuyi <[email protected]>
Cc: Mel Gorman <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
Reported-by: kernel test robot <[email protected]>
Cc: Mike Kravetz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
The devm_clk_get_enabled() helper:
- calls devm_clk_get()
- calls clk_prepare_enable() and registers what is needed in order to
call clk_disable_unprepare() when needed, as a managed resource.
This simplifies the code, the error handling paths and avoid the need of
a dedicated function used with devm_add_action_or_reset().
Based on my test with allyesconfig, this reduces the .o size from:
text data bss dec hex filename
12843 4804 64 17711 452f drivers/rtc/rtc-ti-k3.o
down to:
12523 4804 64 17391 43ef drivers/rtc/rtc-ti-k3.o
Signed-off-by: Christophe JAILLET <[email protected]>
Link: https://lore.kernel.org/r/601288834ab71c0fddde7eedd8cdb8001254ed7e.1661329498.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Alexandre Belloni <[email protected]>
|
|
The devm_clk_get_enabled() helper:
- calls devm_clk_get()
- calls clk_prepare_enable() and registers what is needed in order to
call clk_disable_unprepare() when needed, as a managed resource.
This simplifies the code, the error handling paths and avoid the need of
a dedicated function used with devm_add_action_or_reset().
As a side effect, some error messages are not logged anymore, so also use
dev_err_probe() instead of dev_err() in case of error.
At least the error code will be logged (and -EPROBE_DEFER will be filtered)
Based on my test with allyesconfig, this reduces the .o size from:
text data bss dec hex filename
9025 2488 128 11641 2d79 drivers/rtc/rtc-jz4740.o
down to:
8267 2080 128 10475 28eb drivers/rtc/rtc-jz4740.o
Signed-off-by: Christophe JAILLET <[email protected]>
Acked-by: Paul Cercueil <[email protected]>
Link: https://lore.kernel.org/r/af10570000d7e103d70bbea590ce8df4f8902b67.1661330532.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Alexandre Belloni <[email protected]>
|
|
The devm_clk_get_enabled() helper:
- calls devm_clk_get()
- calls clk_prepare_enable() and registers what is needed in order to
call clk_disable_unprepare() when needed, as a managed resource.
This simplifies the code, the error handling paths and avoid the need of
a dedicated function used with devm_add_action_or_reset().
That said, mpfs_rtc_init_clk() is the same as devm_clk_get_enabled(), so
use this function directly instead.
This also fixes an (unlikely) unchecked devm_add_action_or_reset() error.
Based on my test with allyesconfig, this reduces the .o size from:
text data bss dec hex filename
5330 2208 0 7538 1d72 drivers/rtc/rtc-mpfs.o
down to:
5074 2208 0 7282 1c72 drivers/rtc/rtc-mpfs.o
Signed-off-by: Christophe JAILLET <[email protected]>
Reviewed-by: Conor Dooley <[email protected]>
Link: https://lore.kernel.org/r/e55c959f2821a2c367a4c5de529a638b1cc6b8cd.1661329086.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Alexandre Belloni <[email protected]>
|
|
A previous commit moved the notifications and end-write handling, but
it is now missing a few spots where we also want to call both of those.
Without that, we can potentially be missing file notifications, and
more importantly, have an imbalance in the super_block writers sem
accounting.
Fixes: b000145e9907 ("io_uring/rw: defer fsnotify calls to task context")
Reported-by: Dave Chinner <[email protected]>
Link: https://lore.kernel.org/all/[email protected]/
Signed-off-by: Jens Axboe <[email protected]>
|
|
Only with the big sqe feature they take 128 bytes per entry, but we
unconditionally advance by 128B. Fix it by using sq_shift.
Fixes: 3b8fdd1dc35e3 ("io_uring/fdinfo: fix sqe dumping for IORING_SETUP_SQE128")
Reported-and-tested-by: [email protected]
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/8b41287cb75d5efb8fcb5cccde845ddbbadd8372.1665449983.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>
|
|
This fixes the shadowing of the outer variable rw in the function
io_write(). No issue is caused by this, but let's silence the shadowing
warning anyway.
Reported-by: kernel test robot <[email protected]>
Signed-off-by: Stefan Roesch <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>
|
|
The msg variants of sending aren't audited separately, so we should not
be setting audit_skip for the zerocopy sendmsg variant either.
Fixes: 493108d95f14 ("io_uring/net: zerocopy sendmsg")
Reported-by: Paul Moore <[email protected]>
Reviewed-by: Paul Moore <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
|
|
Running local task_work requires taking uring_lock, for submit + wait we
can try to run them right after submit while we still hold the lock and
save one lock/unlokc pair. The optimisation was implemented in the first
local tw patches but got dropped for simplicity.
Suggested-by: Dylan Yudaken <[email protected]>
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/281fc79d98b5d91fe4778c5137a17a2ab4693e5c.1665088876.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>
|
|
io_cqring_wake() needs a barrier for the waitqueue_active() check.
However, in the case of io_req_local_work_add(), we call llist_add()
first, which implies an atomic. Hence we can replace smb_mb() with
smp_mb__after_atomic().
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/43983bc8bc507172adda7a0f00cab1aff09fd238.1665018309.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>
|
|
We treat EINPROGRESS like EAGAIN, but if we're retrying post getting
EINPROGRESS, then we just need to check the socket for errors and
terminate the request.
This was exposed on a bluetooth connection request which ends up
taking a while and hitting EINPROGRESS, and yields a CQE result of
-EBADFD because we're retrying a connect on a socket that is now
connected.
Cc: [email protected]
Fixes: 87f80d623c6c ("io_uring: handle connect -EINPROGRESS like -EAGAIN")
Link: https://github.com/axboe/liburing/issues/671
Reported-by: Aidan Sun <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
|
|
Notifications were killed but there is a couple of fields and struct
declarations left, remove them.
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/8df8877d677be5a2b43afd936d600e60105ea960.1664849941.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>
|
|
->mm_account should be released only after we free all registered
buffers, otherwise __io_sqe_buffers_unregister() will see a NULL
->mm_account and skip locked_vm accounting.
Cc: <[email protected]>
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/6d798f65ed4ab8db3664c4d3397d4af16ca98846.1664849932.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>
|
|
Instead of putting io_uring's registered files in unix_gc() we want it
to be done by io_uring itself. The trick here is to consider io_uring
registered files for cycle detection but not actually putting them down.
Because io_uring can't register other ring instances, this will remove
all refs to the ring file triggering the ->release path and clean up
with io_ring_ctx_free().
Cc: [email protected]
Fixes: 6b06314c47e1 ("io_uring: add file set registration")
Reported-and-tested-by: David Bouman <[email protected]>
Signed-off-by: Pavel Begunkov <[email protected]>
Signed-off-by: Thadeu Lima de Souza Cascardo <[email protected]>
[axboe: add kerneldoc comment to skb, fold in skb leak fix]
Signed-off-by: Jens Axboe <[email protected]>
|
|
The function name is missing the letter 'd' in the comment block.
Fix it.
Signed-off-by: Colin Ian King <[email protected]>
Acked-by: Joshua Kinard <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexandre Belloni <[email protected]>
|
|
The regmap abstraction allows us to avoid the private i2c transfer
helpers, and also offers some nice utility functions such as the
regmap_update_bits family.
While at it, simplify the code even more by not keeping track of
->write_enabled: rtc_set_time is not a hot path, so one extra i2c read
doesn't hurt (regmap_update_bits elides the write when the bits are
already as desired).
Signed-off-by: Rasmus Villemoes <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexandre Belloni <[email protected]>
|
|
There's nothing in the data sheet that says writing to one of the time
keeping registers is necessary to start the RTC. It does so at the
stop condition of the i2c transfer setting the WRTC bit:
Upon initialization or power-up, the WRTC must be set to "1" to
enable the RTC. Upon the completion of a valid write (STOP), the RTC
starts counting.
Moreover, even if such a write to one of the timekeeping registers was
necessary, that's exactly what we do anyway just below when we
actually write the given struct rtc_time to the device.
Signed-off-by: Rasmus Villemoes <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexandre Belloni <[email protected]>
|